Tep chi
Tin hoc va Dieu khien hoc,
T.20, S.4
(:?004), 351-354
I' " ,
A
lJOC LlJQ'NG MO TRI;\NG THAI H~ PHI TUYEN
VU
NHU LAN,
VU
CHAN HUNG, vA DANG THANH PHU
Vifn Gong nghf thOng tin
Abstract.
The authors have proposed a fuzzy filtering scheme with the asumption that the mem-
bership functions are triangular. The nature of the proposed method in this article ifi.using fuzzy
logic to find the nonlinear function in the problems of estimating system state. This method is much
more simple than the other ones such as [1,2,3].
Tom
t~t. M9t hroc
do
19C
mo diroc
de
xujit voi gici thiet rang ham thuoc co dang tam giac, Ban
chat cua phuang phap
de
xuat la su dung logic mo tim ham phi tuyen trong bai toan iroc
hrcng
trang thaihe thong. Phuang phap nay don gian han so voi nhieu phuang phap khac nhir [1,2,3].
C6 the trinh bay t6m tiit bai toan iroc
hrcng
trang thai kinh dien
nlnr
sau.
Xet h~ phi tuyen:
Xn+l
=
f(xn)
+
Wn,
(la)
Zn
=
h(xn)
+
V
n
,
(lb)
trong
do
vecta
Xn -
trang thai h~ thong
a
thai diem
n,
Wn -
nhieu triing chuan,
Zn -
vecta
quan sat,
Vn -
nhieu quan sat,
f(.)
va
h(.)
la cac vecta ham phi tuyen cua trang thai.
Gia thiet rKng cac ki vong:
E(xo)
=
xo, (2)
E[(xo - xO)lT
=
Po,
(3)
E[wnl
=
E[(vn)l
= 0,
\In, (4)
E[w
n
vTJ
=
QO'nl,
(5)
[vn vTl
=
RO'nl·
(6)
B9
19C
Kalman
mo
rong cho phep tim dircc uoc
hrong
Xn+l
cua trang thai
Xn+l
trim co
sa
cac quan sat
Zn
(n
=
0,1,2, ).
Neu tinh phi tuyen
a
h~ (1) du
trori
thi c6 the khai trien (1) xung quanh trangthai nhir
sau:
f(xn) ~ f(xn)
+ 8~~)
Ix=xn
(xn - xn),
(7a)
h(xn) ~ h(xn)
+
8h;;X)
Ix=xn
(xn - xn).
(7b)
Tir
d6 c6 the viet (1) d iroi dang:
Xn
=
f(Xn-l
+
Kn[zn - h( -Xn+l)],
(8a)
Cho h~
(1),
iroc hrong
tim
diroc co the viet
duoi dang
tong quat:
Xk
=
j(Xk-l)
+
9(Zk' Xk-l).
(9)
352
vO
NHU LAN,
vO
CHAN HUNG, V A DANG THANH PHU
«;
=
PnHn(Rn + H~ PnHn)-l,
P;
=
Fn(P
n
-
l
- KnH~ Pn-dF',[ + Qn,
(8b)
(8c)
voi
Fn = f(xn)
va
Hn = h(xn).
Nhtr v~y iroc hrorig thu duoc chi la gan dung. Dieu nay gQ'i mo cho viec su dung logic
mo. Nhieu tac gici da su dung kha nang nay [1,2,3]' tuy nhien diro
i
day se trmh bay mot
phuong phap
hoan toan
khac, don gian ho'n va
de
su dung hen.
2.
D~T
BAI TOA.N U<1C LUQ'NG
MO
,
a
day
g(.)
chinh la ham mo tci sir sai lech giira tree hrong tirn diroc va trangthai chinh xac
cua he (1). Co the noi rKng khcng co mot
pmrcng
phap nao tir
trtroc
den nay cho phep xac
dinh chinh xac ham nay. Nhir vay co the coi
g(.)
la ham mo. Tir day, gicii phap cho van
de
neu tren la
su
dvng logic mer tim ham g(.) theo quan diem
[4l.
3.
GIAI QUYET BAI TOA.N U<1C LUONG
MO
Cau true bo iroc lirong
me
diroc the hien nhir sau.
Ucc
hrong
19C:
Xk/k
=
Xk/k-l
+
gl(Zk,Xk/k-d·
(lOa)
U'oc
hrong
dtr
bao:
Xk/k-l
=
Xk-l/k-l
+
g2(Zk-l, Xk-l/k-l), k
= 1,2,
gi
=
[gil,gi2, ,ginl
T
,
girc[minir,maxirlcR, r=I,2, :,n,
i=I,2,
vo
i
dieu ki~n ban
dau:
(lOb)
XO/O
=
:ro·
Cac vecto ham
gl (.), g2 (.)
co 2 vecto dau vao
el, .6el va
e2, .6e2
ttro ng
irng
va
tai
thai
diem
k
co:
elk
=
Zk - h(Xk/k-d,
.6elk
=
elk - el(k-l),
e2(k-2)
=
Zk-l - h(Xk-l/k-l),
.6e2(k-l)
=
e2(k-l) - e2(k-2),
(lla)
(llb)
(12a)
(12b)
voi:
eik
=
[eikl' eik2,· , eikml
T
,
.6eik
=
[.6eikl' .6eik2,· , .6eikml
T
,
eikp
C
[min
eikp,
max
eikpl
C
R,
.6eikp
C
[.6 min
eikp,.6
max
eikpl
C
R,
p
= 1,2,
,m.
u6c
LUQNG MOTRANGTHAI HI;; PHI TUYEN 353
Gia Slr r~ng
ca
sa luat momo ta quan he giira cac sai s6
eikp, /:).eikp
va
glr(.), g2r(.)
diroc
cho tai bang 1 va bang 2 theo dinh huang cua tri thirc chuyen gia duo
i
dang ngon ngir tv
nhien: qua nho,
tilu),
hcri nh», truno binh, hcri tun, tun, qua tun.
Dau ra
glr(.)
va
g2r(.)
la 5 tap mo va 7 tap mo tirong irng cling dirci dang ngon ngir tv
nhien nhir tren (xem bang 1 va bang 2).
Phan b6 cac t~p mo tai dau vao va dau ra gia thiet la deu nhau (chua xet den bai toan
t6i iru tham s6 ham thuoc).
Bdng
1. Co
sa
luat
cho
glr(.)
tai
thai diem
k
/:).eIp(k)
qua nho
nh6
trung binh Ian qua Ian
qua nho
qua nho qua nho
nhO
nh6
trung binh
eIp(k)
nhO qua nho
nhO
nh6 trung binh Ian
trung binh nh6 nh6
trung binh
Ian
Ian
Ian nho trung binh Ian Ian qua Ian
qua Ian trung binh Ian Ian qua Ian
qua Ian
Bdng
2. Co sa luat cho
g2r(.)
tai thai diem
(k -1)
/:).e2p(k -
1)
qua nh6 ho
i
trung
hoi
Ian qua
nh6
nhO
binh Ian
Ian
qua qua qua
nhO nhO
ho
i
ho
i
trung
nhO nhO nhO
nhO nhO
binh
nhO qua
nhO nhO
hoi ho
i
trung
ho
i
nhO nh6 nh6 binh Ian
hai
nhO nhO
ho
i
hai trung
hoi
ho
i
e2p(k -
1)
nho
nhO nhO
binh
Ian Ian
trung
nhO
hoi ho
i
trung
ho
i
hoi
Ian
binh nh6
nh6 binh Ian Ian
hai ho
i
hoi trung hoi
hai Ian Ian
.km
nh6 nh6 binh Ian
Ian
Ian hoi
trung
hoi
hoi
Ian Ian
qua
nhO
binh Ian Ian
Ian
qua trung
ho
i
ho
i
Ian Ian
qua qua
Ian binh Ian Ian Ian
Ian
Ham thuoc d6i voi
glr(.)
va
g2r(.)
nen Slr dung tam giac de tinh toan don gian. Nhir
vay co the dinh nghia ham thuoc
voi
3 tham s6
(II
j
,
CL, rI
j
)
cua dau vao
j (j
=
1,2),
(i = 1,2,3,4,5) d6i
voi
glr(.)
va
(lZj'
Cf
j
,
rh)
cua
dau vao
j
(j
= 1,2), (i = 1,2,3,4,5,6,7)
d6i
vo
i
g2r(.).
Ham thuoc co dang: { (_
C )/1"
1
+
X
2Jr 2Jr
f-Lijr(X)
=
1 -
(x -
Cijr)/rijr
o
(; day:
x ~ eikr
hoac
/:).eikr
neu
-lijr::; (x - C
ijr
) ::;
0,
neu 0::;
(x - C
ijr
) ::; rijr,
cac trirong ho p can 10i.
(13)
7 0 ) 0
L
!-Li2r(Yi2r Yi2r
i=l
(
Xk 11k-I)
= 7 ( 0 )
92r Zk-l, -
L
!-Li2r Yi2r
i=l
(15)
354
vO NHlJ LAN, vO CHAN HlJNG, vA BANG THANH PHU
D~u ra
91r (.)
sau phep giai mo theo phirong phap trong tam co dang:
5
L
!-Lilr(Y?lr)Y?lr
91r(ZkXklk-l)
=
-i=-1-::-5
L
!-Lilr(Y?lr)
i=l
(14)
veri
Y?lr -
trong tam tap mo dau ra tlnr
i
cua
glr(.),
Y?2r -
trong tam tap mo dau ra thir
i
cua
92r(.),
va
!-Lilr(Y?lr)
=
sUP{!-Lelr(e~kp)
A
!-LLlelkpr(~e~kp)},
!-Li2r(Y?2r)
=
SUp
{!-Le2
r
(egkp)
A
!-LLle2kpr(~egkp)}·
V
eri t~p
dau vao ro tinh
diroc
truce:
elkp
=
e~kp , ~elkp
=
~e~kp ;
(18)
e2kp
=
egkp , ~e2kp
=
~egkp·
(19)
Nhir v~y da
xac
dinh diroc iroc IUQ'ng trangthai tai cac thai diem
k
theo (10), (14) va
(15). Liru
y
rKng trong tnrorig hop e va ~e qua Ion co the phat trien them phan t6i tru h6a
tham s6 ham thuoc. Phan nay se diroc tiep tuc xem xet them trong tirong lai.
(16)
(17)
4.
KET LUAN
Bai bao da
ae
xuat each giai quyet tuang d6i don gian bai toan iroc hrcng trangthai h~
phi tuyen dua tren ca
sa
logic mo. Giai phap nay khac hiin veri cac giai phap [1,2,3]. Phuong
phap tren con co the S11 dung khi nhieu
w(n)
va
v(n)
trong mo hinh (1) khong phai la trang
chuan. Truong hop nay cac phuong phap iroc IUQ'ngkinh dien khong giai quyet diroc.
TAl Ll:¢U THAM KHAO
[1]
Fernandez S. A., Lopez C. A., and Alzola
J.
R.,
A fuzzy-controlled Kalman filter applied
to stereo-visual tracking schemes, Signal Processing
83
(2003) 101-120.
[2]
Kobayeshi K., Cheok K., and Watenable K., Estimation of the absolute vehicule speed
using fuzzy logic rule-based Kalman filter, American Control Conj., Seattle, 1985, 3086-
3090.
[3]
Simon D., Fuzzy estimation of DC motor winding currents, North American Fuzzy Infor-
mation Processing Society Conj., NewYork, 1999, 859-863.
[4] VU Nhir Lan, VU Chan Hung, va Dang Thanh Phu, Phuong phap mci mo hinh mo h~
dong h9C chira bat dinh, Top chi Khoa h9C va Cong ngh~
40
(so DB) (2002) 115-110.
Nluin
bdi ngay 10- 12 - 2003
. S.4
(:?004), 351-354
I' " ,
A
lJOC LlJQ'NG MO TRI;NG THAI H~ PHI TUYEN
VU
NHU LAN,
VU
CHAN HUNG, vA DANG THANH PHU
Vifn Gong nghf thOng tin
Abstract.
The. dang tam giac, Ban
chat cua phuang phap
de
xuat la su dung logic mo tim ham phi tuyen trong bai toan iroc
hrcng
trang thai he thong. Phuang phap nay don